A comprehensive update on the application of high-throughput fluorescence imaging for novel drug discovery

Abstract

Introduction:

High-throughput fluorescence imaging (HTFI) is revolutionizing drug discovery by enabling rapid and precise detection of biological targets and cellular processes. Recent advances in fluorescence imaging technologies now provide unprecedented sensitivity, resolution, and throughput. Integration of artificial intelligence (AI) and machine learning (ML) into HTFI workflows significantly enhances data processing, aiding in hit identification, pattern recognition, and mechanistic understanding.

Areas covered:

This review outlines recent technological developments, integration strategies, and emerging applications of HTFI. It emphasizes HTFI’s role in phenotypic screening, especially for complex diseases such as cancer, neurodegenerative disorders, and viral infections. Additionally, it highlights advances in 3D culture systems, organoids, and organ-on-a-chip technologies, which facilitate physiologically relevant testing, improved predictive accuracy, and translational potential, alongside innovative molecular probes and biosensors.

Expert opinion:

Despite its advancements, HTFI faces ongoing challenges, including data standardization, integration with multi-omics approaches, and scalability of advanced models. However, recent progress in organoid and 3D modeling technologies has enhanced the physiological relevance of HTFI assays, complemented by sophisticated AI and ML-driven data analysis techniques.

1. Introduction to high-throughput fluorescence imaging in drug discovery

High-throughput fluorescence imaging (HTFI) enables direct visualization of biological targets and assessment of cellular processes and as such has become an indispensable tool in the field of drug discovery, offering transformative capabilities for screening and characterizing large compound libraries with efficiency, precision, and scale [1]. The advent of HTFI has enabled researchers to perform thousands of assays in parallel with high sensitivity and reproducibility, vastly increasing throughput and reducing the time needed to identify viable drug candidates. This shift has been facilitated by advances in fluorescent imaging technologies, automation, artificial intelligence, and high-content analysis algorithms, all of which enable detailed interrogation of cellular and molecular dynamics under various treatments [2, 3]. The scope of HTFI extends from early-stage discovery and target identification to lead optimization and preclinical evaluation, making it integral across the entire drug development pipeline.

The core of HTFI’s utility is its ability to visualize and quantify specific molecular and cellular responses using sensitive fluorescence-based detection. Fluorescent probes can be designed to bind or be fused to specific cellular targets, allowing researchers to observe interactions, pathway activations or suppressions, and phenotypic changes that result from drug exposure or other environmental changes [4–6]. Techniques such as multicolor fluorescence, fluorescent lifetime imaging, and high-content imaging expand this capability further, allowing for the simultaneous tracking of multiple targets or cellular events. Fluorescence probes are particularly useful as they enable direct endogenous target labeling without the need for genetic engineering with the advantage of cell permeability, low cytotoxicity, and high quantum yield.

One of the more common applications of HTFI in drug discovery is in phenotypic screening, a strategy that identifies hit compounds based on observable effects in cellular models rather than specific molecular targets. This unbiased approach enables project teams to uncover novel therapeutic pathways and potential off-target effects that might be overlooked by target-based assays. HTFI is particularly valuable in phenotypic screening for diseases with complex pathologies, such as cancer, neurodegeneration, and viral infections [7]. For example, in cancer research, HTFI can be used to monitor tumor responses to a wide range of compounds in real time, providing insights into apoptosis, cell cycle progression, and cytotoxicity. Similarly, in virology, fluorescence-based assays are used to detect viral entry, replication, and inhibition of processes in infected host cells, expediting the identification of antivirals with therapeutic potential.

Another significant advantage of deploying HTFI is its applicability to three-dimensional (3D) cell cultures, organoids, and tissue models, which better mimic the in vivo environment than the traditional 2D cell cultures [8]. Academic and industry drug discovery centers have made massive investments in time and architecture into advanced imaging platforms that enable high-throughput analysis of these 3D systems, providing more physiologically relevant data on drug effects, including tissue-specific toxicity and complex cellular interactions. This is particularly important in areas such as oncology, where tumor organoids and patient-derived spheroids offer a closer approximation of human cancer biology. Additionally, organ-on-a-chip technologies combined with HTFI allow for high-throughput, tissue-specific drug testing in complex microenvironments, improving the predictive accuracy of preclinical studies.

The integration of artificial intelligence (AI) and machine learning (ML) into HTFI has further enhanced its analytical power and ability to handle the extremely large datasets that a typical HTFI screen produces [9]. AI-driven algorithms can process the large datasets generated by high-content screening more quickly and accurately than human analysts, extracting features and identifying patterns that inform drug efficacy and safety profiles [10]. For example, deep learning models trained on HTFI data can recognize subtle morphological changes in cells or organelles that indicate a compound’s mechanism of action or potential side effects. Moreover, AI-driven automation reduces the need for manual intervention, increases throughput, and allows for more standardized data interpretation, facilitating the reproducibility of screening results across different studies.

While HTFI has advanced drug discovery projects, challenges in its employment remain, particularly around data management, standardization, and integration with other omics approaches [11]. The high-dimensional data generated by HTFI require robust bioinformatics pipelines to manage, analyze, and store. Standardization across laboratories is essential to ensure reproducibility and reliability of results, especially when comparing results from diverse platforms and models. Additionally, the integration of HTFI data with multi-omics, such as genomics, transcriptomics, and proteomics, offers a holistic view of drug action but requires sophisticated computational tools to integrate and interpret this multidimensional data.

By enabling high-throughput phenotypic screening, providing physiologically relevant data from complex 3D models, and incorporating AI-driven analysis, HTFI facilitates the identification of promising therapeutic compounds more efficiently than traditional methods. As technologies advance, HTFI is likely to become even more integrated into the drug discovery process, with continued innovations in imaging, automation, and data integration enhancing its potential to accelerate and refine drug development. This review explores the evolution and impact over the past three years of HTFI from academic and industry perspectives, advancements in image and data analysis pipelines, the integration of multiomics and data repositories, and future directions for hardware and software development.

2. Applications of high-throughput fluorescence imaging in drug discovery

2.1. Phenotypic screening

HTFI excels at phenotypic screening by enabling unbiased characterization of observable cellular changes without predefined molecular targets. Unlike traditional target-based screening, this approach is less biased, allowing the cells or organisms themselves to reveal the most relevant targets necessary for achieving a desired therapeutic outcome [12–14]. Such screening is crucial in identifying novel targets and pathways for diseases with complex pathologies, such as cancer, neurodegenerative diseases, and infectious diseases. Several reviews have discussed the power of morphological profiling in response to small molecules using multi-parameter cellular assays [3]. Recently, morphological profiling for Alzheimer’s disease using a cell morphology-based drug screen centered on the risk gene, sorl1 (which encodes the protein sorla). Increased Alzheimer’s disease risk has been repeatedly linked to variants in sorl1. An automated morphological profiling method (cell painting) was adapted to neural progenitor cells and used to determine the phenotypic response of sorl1 −/− neural progenitor cells to treatment with compounds from an approved drug library [15].

Indeed, cell painting, a high-content morphological profiling technique, has gained prominence in driving phenotypic screening campaigns. This approach utilizes multiplexed fluorescent dyes to label various organelles and cellular structures, creating comprehensive morphological fingerprints of cells. By capturing phenotypic responses across diverse cellular components, cell painting enables the identification of compounds with distinct or unexpected mechanisms of action. The technique’s capacity for unbiased, large-scale profiling makes it invaluable for both drug discovery and toxicity assessment. Notably, cell painting has been integrated into numerous high-throughput platforms to enhance hit identification and mechanistic studies, accelerating the drug development pipeline for targets. Researchers recently adapted the cell painting technique to identify the phenotypic signatures of PROTACS, blending the high-content imaging method with chemical clustering and model predictions to identify signatures that would not be recognized were the individual PROTAC components screened [16].

Innovative HTFI methodologies continue to push the boundaries of phenotypic screening and address bottlenecks in the platform. Junfang et al. (2023) introduced a high-throughput and high-content drug screening platform combining superhydrophobic microwell array plates (SMAP) with protein-retention expansion microscopy (ProExM) [17]. This system achieved super-resolution imaging of microtubules at 68 nm within three hours, bridging the gap between high-throughput capabilities and detailed structural analysis. Similarly, Fricke et al. (2022) showcased HTFI’s utility in screening modulators of plasma membrane rafts, which play pivotal roles in cellular signaling and membrane dynamics [18]. By targeting these complex cellular structures, HTFI facilitated the identification of compounds with potential to modulate disease-relevant pathways.

Unbiased phenotypic screens have also proven effective in rare disease research. Afshin et al. (2024) developed a high-content screening assay that identified BCH-HSP-C01 as a lead compound capable of correcting ATG9A trafficking defects in AP-4 deficiency, a rare hereditary spastic paraplegia [19]. Screening over 28,000 small molecules against patient-derived fibroblasts and iPSC-derived neurons using automated image analysis alongside transcriptomic and proteomic data elucidated potential mechanisms of action of this proof-of-concept therapeutic candidate.

Higher-order chromatin structures play critical roles in gene regulation and cell identity but are challenging to analyze or visualize at the single-cell level. A machine learning-based computational method, extracting physical characteristics from images of chromatin structures (EPICS), was developed to process high-resolution 3D chromatin imaging data from techniques such as 3D-EMISH and 3D-SIM. Leveraging HTFI data of chromatin domains into physical representations, EPICS generated direct 3D representations of chromatin structures, identified chromatin domains, and determined their open or closed states, revealing physical features defining chromatin domain states. In this manner, EPICS provides a valuable tool for spatial genomics and high-resolution chromatin structure analysis [20].

As another target of HTFI, mitochondrial morphologies vary across cell types and developmental stages, reflecting functional needs, but their quantification in complex systems remains challenging. The mitochondrial cellular phenotype (MitoCellPhe) tool was developed to provide high-throughput, quantitative analysis of mitochondrial morphology by generating highly accurate skeletons of mitochondria structures, generating 24 parameters for comprehensive structural assessment in both single cells and cell clusters. Using MitoCellPhe, networks of mitochondria in healthy fibroblasts and fragmented morphologies were validated and morphological differences between healthy and diseased states were identified in both cell types. This tool enables detailed insights into mitochondrial dynamics previously unavailable to researchers [21].

2.2. 3D models and organoids

High-throughput drug screening has long been hindered by the limitations of 2D cell cultures which fail to replicate the complexity and architecture of more relevant 3D tissues. The emergence of 3D models and organoids has changed this landscape, offering more physiologically relevant platforms that bridge the gap between in vitro experimentation and in vivo outcomes and integrating naturally with current HTFI platforms. One of the most notable recent breakthroughs has been the development of automated 3D imaging pipelines that integrate machine learning to analyze complex cellular structures. For instance, the creation of kidney organoids from human pluripotent stem cells has provided a robust model for nephrotoxicity screening and disease modeling. Haruka et al. (2024) leveraged this technology to model acute kidney injury induced by cisplatin, uncovering imatinib as a potential therapeutic agent capable of mitigating drug-induced damage [22].

The application of 3D models also extends to metastasis research, where tumor cell dissemination and migration can be studied more effectively. Hossam et al. (2022) embedded osteosarcoma spheroids in aligned collagen matrices, developing assays that not only quantify cell movement but also enable high-throughput evaluation of potential anti-metastatic therapies [23]. This model preserves the integrity of the extracellular environment, providing insights into the mechanisms driving metastasis. In hepatocellular carcinoma (HCC) research, Sang-Yun et al. (2022) optimized aggregated spheroid models (3DASM) to reflect tumor heterogeneity more accurately than 2D cultures [24]. These spheroids support immunofluorescence staining and extracellular matrix integration, allowing for selective drug screening that mirrors in vivo conditions. By assessing the efficacy of FDA-approved drugs like sorafenib, this approach underscores the value of 3D platforms in enhancing the precision of high-throughput cancer drug screening.

Glioblastoma research has also benefited from 3D bioprinting technologies, with the development of neurovascular unit models that recapitulate the brain microenvironment. Yen-Ting et al. (2024) designed systems incorporating patient-derived glioblastoma cells alongside astrocytes, pericytes, and endothelial cells, providing a dynamic model for tumor growth and vascularization [25]. Screening of chemotherapeutics within these platforms has validated their relevance for high-throughput applications and personalized treatment approaches. Microfluidic hanging drop systems offer another dimension of control, enabling precise manipulation of spheroid geometry and size. Ganguli et al. (2021) demonstrated that glioblastoma and xenograft models cultivated in these systems retain their genomic characteristics, allowing for real-time drug response monitoring and advancing personalized oncology initiatives [26].

Beyond organoids, advances in 3D imaging flow cytometry have enabled scalable analysis of spheroids and other multicellular structures. Minato et al. (2024) utilized light-sheet microscopy combined with microfluidic devices to facilitate rapid, high-throughput imaging of hydrogel-embedded cells, addressing the long-standing trade-off between imaging speed and resolution [27]. While spatial resolution may be lower, the ability to analyze large numbers of samples in parallel represents a significant step forward for fields like cancer biology and tissue engineering. Similarly, human induced pluripotent stem cell spheroids are increasingly used to model early development and tissue regeneration. Haneen et al. (2021) employed convolutional neural networks to analyze live phase contrast imaging of spheroids, predicting culture conditions and tracking differentiation in real time and enhancing the scalability and sensitivity of related protocols [28].

Automated tools are essential for managing the vast datasets generated by these technologies. To streamline spheroid analysis, Akshay et al. (2022) introduced SpheroScan, a deep learning framework offering rapid and reliable detection of spheroids across different imaging platforms [29]. This not only accelerates drug discovery workflows but also democratizes access to 3D analysis by providing user-friendly, web-based interfaces. Advancements in segmentation algorithms such as Acu2Net have also refined organoid imaging, providing superior accuracy in analyzing cancer phenotypes, particularly organoids. Shudi et al. (2023) developed the Acu2Net tool by leveraging deep learning to overcome challenges associated with large-scale image analysis, streamlining workflows and increasing the throughput of organoid-based drug screens compared to manual analysis methods (i.e., ImageJ segmentation) [30].

2.3. Disease modeling

The development and integration of innovative tools and HTFI approaches continues to drive discoveries and enable more accurate modeling of disease states. The development of innovative tools like Phaser-TRIM, a genetically encoded biosensor, facilitates in vivo imaging of histone methylation dynamics by visualizing H3K9me3 modifications through phase-separated droplets, allowing real-time monitoring of epigenetic changes during neural stem cell differentiation and cancer progression [31]. Similarly, the pursuit of novel metastasis therapies has driven advancements in imaging flow cytometry, with artificial intelligence-enhanced approaches accelerating the identification of factors influencing cortical cell polarity in melanoma cells, thus improving drug screening efficiency and reproducibility [32].

Neurodegenerative research has similarly benefited from technological innovations in HTFI, with automated high-throughput neurite outgrowth assays utilizing fluorescence imaging and live-tracking to quantify neuronal growth, supporting investigations into Parkinson’s disease (PD) and other neurodegenerative disorders [33, 34]. Induced pluripotent stem cell (iPSC)-derived cortical and motor neurons have provided scalable platforms for identifying neurotoxic compounds and screening therapeutic candidates for PD, with promising kinase inhibitors such as BX795 demonstrating potential to restore autophagy and normalize protein synthesis [35]. Patient-derived models of tauopathies, such as those carrying the MAPT P.A152T mutation, have also enabled automated screening of kinase inhibitors to mitigate tau phosphorylation and neuronal death [36]. Similarly, the need to address cellular senescence and its implications in aging and disease prompted the development of FAST (Fully Automated Senescence Test), a high-throughput, image-based method that quantifies SA-β-gal activity and proliferation arrest, providing robust single-cell assessments and advancing drug discovery efforts targeting senescence [37].

Inflammatory disease research has also advanced through high-throughput screening platforms targeting the NLRP3 inflammasome. Image-based assays employing fluorescently tagged ASC proteins in macrophages facilitate the identification of novel inhibitors, contributing to the development of therapies for neurodegenerative and autoimmune conditions [38]. On a broader scale, tools like HIIDDD (High-Throughput Immune Cell DNA Damage Detection) have enhanced DNA damage quantification in immune cells, providing insights into cancer progression and immune aging [39].

Expanding beyond mammalian cells, bacterial research has been transformed by high-throughput transposable fluorescent promoter probes, which enable the identification of heterogeneously expressed genes in bacteria with high spatial and temporal resolution [40]. Furthermore, fungal antibiotic susceptibility testing has been streamlined through image-based assessments, allowing for the rapid evaluation of fungal responses to antibiotics [41].

2.4. Drug safety and toxicity

Fluorescence imaging assays are becoming essential tools for evaluating drug safety and toxicity by enabling precise visualization and quantification of cellular and molecular changes. Emerging research highlights the link between environmental toxicants, particularly insecticides and endocrine-disrupting chemicals (EDCs), and adverse health outcomes mediated through epigenetic and cellular pathways. Insecticides such as imidacloprid have been shown to alter DNA methylation during critical stages of embryogenesis and cell differentiation, influencing cell processes like apoptosis and the cell cycle. High-content imaging studies in mouse embryonic stem cells reveal that these methylation changes offer a promising route for developmental toxicity evaluation [42]. Recent advances using 3D microtissue models, quantitative imaging, and machine learning have also enabled precise assessment of estrogenic EDCs by analyzing microtissue organization and luminal volume [43]. Using this fully automated tool with a 3D culture system of MCF7 breast cancer microtissues, researchers achieved ML model training AUC values of 0.95 for E2 and PPT exposure while deep learning software characterized gland lumen formation successfully.

In genetic toxicology, an automated high-throughput in vitro micronucleus assay has been developed to assess small molecule genotoxic potential. Integrating acoustic dosing and confocal imaging, the assay evaluates micronuclei, cytotoxicity, and cell-cycle profiles by incorporating kinetochore labeling and detection of foci formation. By integrating machine learning and Bayesian classifiers, researchers achieved 95% accuracy in classifying compounds while reducing analysis time by 80% and minimizing human bias [44].

Similarly for developmental toxicity, the embryonic stem cell test (EST) remains a validated in vitro system but faces adoption hurdles due to long assay durations and low throughput. A next-generation EST workflow addresses these issues by reducing assay time, incorporating homogeneous viability assays, and using high-content imaging with flow cytometry to detect embryotoxicants and provide a scalable platform for industrial toxicity screening [45].

Respiratory toxicity assessment has also seen innovation with a high-content imaging assay that quantifies occludin in A549 lung epithelial cells to evaluate barrier integrity [46]. This assay accurately identified 90% of compounds with respiratory risks and 100% without, aligning with non-clinical and clinical profiles. PBPK modeling linked in vitro findings to in vivo lung pathology, enhancing translational risk assessment for inhaled therapeutics.

3. Advanced imaging and computational tools

3.1. High-Throughput and Multiplexed Imaging Platforms

As fluorescence imaging continues to evolve as a cornerstone of high-throughput screening (HTS) in drug discovery, new methodologies and platforms enable more precise, multiplexed analyses of cellular and molecular processes with higher-throughput and data density. One notable advancement is MAC-SEQ, a cost-efficient, ultra-high-throughput RNA-seq workflow designed for 384-well plate formats [47]. This technique facilitates detailed time-course studies, supporting both 2D and 3D cell cultures. With the ability to identify over 10,000 expressed genes per well at sequencing depths of one million reads, MAC-SEQ seamlessly integrates transcriptomic profiling with high-content imaging. By allowing visual detection of cell perturbations at scale and combining this with transcriptomic datasets, the platform broadens the scope of HTS applications and enhances our capacity to monitor complex cellular responses to drug candidates.

Complementing transcriptomic advances, a new high-throughput immunofluorescence (IF)-based assay optimizes the analysis of Smad phosphorylation (pSmad) signaling pathways and enables studies at the single cell level using HTFI [48]. Using BMP-2 and TGF-β1 as signaling cues, the assay applies high-content imaging in 96-well microplates to investigate pathway-specific inhibitors, matching the detection limit of western blotting while offering broader applicability to cytoskeletal and surface marker analyses. These findings offer valuable perspectives for single-cell mechanistic studies and provide potential applications in cancer immunotherapy, significantly expanding the scope of high-throughput signaling pathway analyses.

3.2. Automated Imaging and Dynamic Live-Cell Analysis

Automation has driven significant progress in imaging-based drug discovery, particularly in live-cell and tissue studies. The Mantis platform exemplifies this trend by integrating oblique light-sheet fluorescence microscopy with remote-refocus, label-free microscopy for dynamic 4D imaging [49]. This system, capable of imaging 20 cell lines every 15 minutes over 7.5 hours, greatly enhances the ability to study viral infections, cellular dynamics, and organelle behavior in a high-content platform. Paired with shrimpy, the paired open-source analysis software, the platform streamlines single-cell phenotyping and segmentation, offering an efficient, scalable approach to imaging large datasets with applications from tracking disease-driven cellular reprogramming to tissue pathology and cellular differentiation studies. A parallel innovation comes from applying fisheye-like transformation method, designed to improve single-cell phenotyping by emphasizing local microenvironmental data. By prioritizing pixels near the cell of interest, the researchers show that this method surpasses conventional deep learning models in accuracy, facilitating more reliable image-based phenotyping across diverse datasets and better integrating spatial context into imaging workflows [50].

3.3. Imaging cytometry

Relatedly, recent advancements in fluorescence image-based cytometry have significantly enhanced the capabilities of non-invasive cell analysis, with profound implications for fields like immuno-oncology, respiratory diseases, and spatial biology. Patel et al. (2021) introduced a non-invasive imaging-based cytometry method to evaluate natural killer (NK) cell-mediated cytolysis in a high-throughput format [51]. This innovative platform facilitates the screening of immune-modulating drugs by quantifying NK cell cytotoxic activity, addressing a critical need in cancer immunotherapy development.

In the context of acute respiratory distress syndrome (ARDS), Oleksii et al. (2021) developed a novel image-based cytometry method to assess endothelial barrier function [52]. Their approach combines the xPERT assay with the Celigo image cytometer, enabling visualization and quantification of endothelial monolayer permeability via fluorescence imaging. This method addresses the limitations of traditional techniques such as transepithelial electrical resistance (TEER) and tracer permeability assays by providing kinetic, dose-dependent measurements of endothelial disruption induced by thrombin, TNF-α, and lipopolysaccharides (LPS). By reducing assay variability and enhancing precision, this platform contributes to high-throughput drug discovery and offers new insights into ARDS pathophysiology. Similarly in spatial biology, Zunming et al. (2022) developed a high-throughput technique to correlate individual nonadherent cell positions with their three-dimensional (3D) imaging features at single-cell resolution [53]. Their method integrates a custom 3D imaging flow cytometer capable of capturing images of 500 to 1,000 cells per second with a robotic cell placement platform that dispenses cells in a first-in, first-out manner. By employing marker beads and DNA sequencing software for error detection, the method achieves highly accurate single-cell analysis. Experiments with human cancer cell lines successfully generated 3D side scattering, fluorescence, and 2D transmission images of approximately 100,000 cells in under ten minutes, bridging single-cell image analysis with molecular profiling.

3.4. Emerging Technologies in Imaging Flow Cytometry and Microfluidics

Recent advancements in imaging flow cytometry and microfluidics have further expanded the potential of fluorescence imaging in drug discovery. A sheathless microfluidic imaging flow cytometer with stroboscopic illumination was developed to achieve blur-free fluorescence detection and subcellular localization at ultra-high throughput [54]. This platform achieves microscopy-level resolution, detecting structures as small as 500 nm with analytical throughputs exceeding 60,000 cells/sec. Applications include investigating phase-separated compartments and screening rare events, overcoming traditional trade-offs between resolution and throughput. High-content screening methods have similarly progressed through microfluidic and imaging innovations. A new HTFI platform integrates microfluidic chips with high-speed imaging, achieving 30 Hz video rates at 0.8 μm resolution that facilitates simultaneous testing of multiple drugs, demonstrating its utility in cardiomyocyte assays for cardiotoxicity evaluation [55]. Further advancements include real-time segmentation and tracking of live cells, 3D mitochondrial imaging for cardiovascular disease models, and mid-infrared chemical imaging using plasmonic metasurfaces for non-destructive, label-free cellular imaging [56–58]. These cutting-edge technologies promise to enhance metabolic phenotyping and live-cell imaging, contributing to next-generation drug discovery.

4. AI and machine learning

The integration of artificial intelligence (AI) and machine learning (ML) into fluorescence imaging workflows has revolutionized drug discovery by enhancing data analysis, automating image processing, and enabling the extraction of complex biological insights from high-content screening (HCS) datasets. These computational tools address longstanding challenges such as data variability, labor-intensive image segmentation, and the need for reproducible, high-throughput analysis. A study by Shapira et al. (2024) highlights the growing adoption of HCS integrated with AI and ML to streamline hit detection and improve data reliability [59]. To address this need, the authors propose an automated hit detection framework utilizing a statistical model and introducing a “virtual plate” method to rescue data from compromised wells, thus enhancing the robustness of HCS campaigns and maximizing the number and quality of hits obtained.

The application of ML to fluorescence imaging extends beyond basic hit detection. Daniel et al. (2022) demonstrated the power of machine learning-driven trans-channel fluorescence prediction to infer biologically relevant markers, such as phosphorylated tau (AT8-ptau) signals, from existing YFP-tau and DAPI channels in Alzheimer’s disease studies [60]. This innovative approach identified active compounds overlooked by conventional analysis, underscoring the potential of ML to maximize the value of archival datasets and refine compound prioritization. Similarly, Luke et al. (2022) employed ML-based image analysis to quantify protein inclusion formation in ALS models, identifying a synergistic small-molecule combination therapy that reduced mutant SOD1-associated cell death [61].

Deep learning (DL) techniques have further refined fluorescence imaging workflows by facilitating functional screening of small molecules with more sensitive and robust measurements. Gyuwon et al. (2022) present a DL-based framework that leverages live-cell microscopy images to classify mesenchymal stem cells (MSCs) according to differentiating stress-enduring (MUSE) markers [62]. The optimized DenseNet121 model achieved high classification accuracy (AUC of 0.975), providing a non-invasive, high-throughput QC strategy for MSC biomanufacturing. Phenotypic profiling methods such as cell painting have also benefited significantly from AI-driven analysis. Floriane et al. (2024) describe the use of ML models to integrate cell painting data with structural and molecular information, enabling the prediction of biological endpoints and modes of action [63]. Despite the promise of cell painting assays, challenges related to batch effects and protocol standardization remain. Addressing these issues through advanced pre-processing strategies and fair data practices is crucial for enhancing the reproducibility and predictive accuracy of cell painting-based drug discovery pipelines.

Advances in real-time cell tracking algorithms have accelerated high-throughput cellular imaging, addressing key segmentation and analysis challenges. Ting-Chun et al. (2023) introduce the FACT (Fast and Accurate Real-Time Cell Tracking) algorithm, which employs GPU-accelerated segmentation and Gaussian-mixture-model-based cell linking to achieve segmentation speeds up to 93.5 times faster than conventional methods [56]. This technology facilitates rapid single-cell profiling, exemplifying the transformative potential of AI in fluorescence imaging. AI has also enabled significant improvements in microscopy image segmentation, a critical component of fluorescence imaging analysis and a source of potential human bias. Garcia Santa et al. (2022) describe a hybrid approach combining traditional computer vision with deep learning to segment autophagy events in microscopy images [64]. By training a DL model on weak labels generated by classical methods, the authors achieved a 25% improvement in segmentation accuracy while minimizing labeling costs, demonstrating the scale of improvements to microscopy workflows brought about by AI.

5. Novel molecular probes and reporters

Recent advances in fluorescent reporter technology have expanded the toolkit for high-throughput screening (HTS), enhancing sensitivity, specificity, and the ability to probe complex biological processes. These innovations leverage molecular engineering, supramolecular chemistry, and novel biosensor designs to facilitate drug discovery and mechanistic studies across diverse therapeutic areas. One notable area of development focuses on lysosomal pH regulation, which plays a critical role in neurodegenerative diseases. The FIRE-PHLY fluorescent lysosomal pH biosensor was designed to monitor and modulate lysosomal acidity in differentiated neuroblastoma cells and enable HTFI phenotypic screening [65]. Screening efforts identified two compounds, OSI-027 and PP242, as effective acidifiers, demonstrating the potential for lysosomal-targeted therapies in neurodegeneration and providing a rich dataset for others to explore novel regulatory pathways of lysosomal pH regulation.

Advances in monitoring membrane protein dynamics recently led to the creation of a SNAP-tag and near-infrared (NIR) imaging-based assay for tracking GPCR degradation [66]. By quantifying receptor half-lives in HEK293 cells, researchers applied an HTFI approach to identify proteasome-dependent degradation pathways, offering a scalable platform for high-throughput GPCR stability studies and expansion of the toolset available for targeting membrane proteins. Similarly, oxidative stress, a key factor in numerous diseases, can now be visualized through a split-MutT protein system [67]. This fluorescent tag system forms foci in response to 8-oxo-dGTP accumulation, providing a real-time readout of nucleotide oxidation. The enhanced sensitivity of this assay allows for high-throughput screening of MTH1 inhibitors, shedding light on oxidative damage pathways and potential therapeutic targets.

In infectious disease research, near-infrared (NIR) fluorescent probes such as LXMB have proven essential for tracking β-lactamase (BlaC) activity in Mycobacterium tuberculosis [68]. LXMB not only facilitated the discovery of tannic acid as a BlaC inhibitor but also enabled live-cell imaging of M. tuberculosis within macrophages. The development of orange wavelength UNAG–ligand pairs further enhance fluorescence microscopy for bacterial systems that are oxygen-sensitive [69]. Using HTS, a benzothiazole-based ligand with high affinity for UNAG was identified, improving fluorescence intensity and red-shifting the emission. This fluorophore facilitates imaging in anaerobic environments, broadening the applications of fluorescence microscopy to previously challenging contexts.

Further expanding the capabilities of fluorescent sensors, a supramolecular tandem assay was developed to detect histone deacetylase 1 (HDAC1) activity [70]. By exploiting SC4A–LCG host-guest interactions, this sensor identified ginsenoside RK3 as a novel HDAC1 inhibitor, emphasizing the role of supramolecular chemistry in drug discovery. Fluorogenic substrates, such as the two-photon F8 probe for cytochrome P450 3A4 (CYP3A4), provide another reporter substrate for HTS screening of CYP modulators [71]. F8’s ability to visualize CYP3A4 activity in living cells and tissues aids in evaluating drug metabolism and screening inhibitors, addressing critical needs in pharmacokinetics and toxicity studies. BODIPY-based probes have also been developed for noninvasive and dynamic imaging of carboxylesterases (CES). BDPn2-CES demonstrated 182-fold fluorescence enhancement within 10 minutes of dosing, enabling real-time imaging of CES activity in living cells. This probe also facilitated HTS, identifying WZU-13 as a potent CES inhibitor, providing a new avenue for CES-related disease diagnostics and therapy [72]. Similarly, S-acylation, a post-translational modification, has been explored using palmitoyl-CoA mimics to visualize DHHC palmitoyl transferase activity [73]. Tracking autophagic flux in live cells has also been optimized through the introduction of HCFP, a hydrophilic, cell-permeable sensor. Its ratiometric pH-sensitive fluorescence allows for precise monitoring of autophagic vesicles, supporting the identification of autophagy modulators and expanding our understanding of autophagy-related diseases [74].

6. Case studies and emerging applications

6.1. Antiviral drug discovery

The COVID-19 pandemic has underscored the urgency of developing antiviral drugs, and fluorescence-based approaches have been pivotal in accelerating this process. A significant advancement in the field has been the development of high-content imaging platforms for identifying SARS-CoV-2 therapeutics. One such platform demonstrates the power of integrating multiple orthogonal datasets including cellular dyes, immunostaining, and advanced image analysis to assess antiviral activity across SARS-CoV-2 variants, including Omicron BA.5 and XBB.1.5, as well as SARS-CoV and human coronavirus 229E. Screening of approximately 900,000 compounds identified candidates with broad-spectrum activity while minimizing host cell toxicity, demonstrating the platform’s capacity to expedite drug discovery and pandemic preparedness [75]. Similarly, fluorescence-based cytometry assays have proven indispensable in large-scale antiviral screening. St. Clair et al.’s imaging cytometry assay for SARS-CoV-2 exemplifies this, providing a robust and reproducible approach for rapid screening of therapeutic compounds while eliminating days off of the standard processing time for antiviral screening [76]. This work complements Chiu’s dual-reporter assay, which enables simultaneous analysis of viral entry and host-cell toxicity, streamlining high-content imaging and enhancing the assessment of large compound libraries against SARS-CoV-2 [77].

Targeting the SARS-CoV-2 spike protein and its interaction with ACE2 has been a focus of SARS-CoV-2 therapeutic development with HTFI application. A BSL-2 compatible assay system using replication-competent vesicular stomatitis virus expressing SARS-CoV-2 spike protein (VSV-eGFP-SARS-CoV-2) facilitated the identification of six natural products with potent inhibitory effects on viral entry [78]. Through this work, scillaren A and 17-aminodemethoxygeldanamycin emerged as inhibitors of spike-S1 binding to ACE2, offering promising leads for further clinical validation. This complements the development of SURF, a bright and reversible fluorogenic reporter used to monitor real-time spike-ACE2 interactions. Through high-throughput screening, SURF identified three natural compounds that inhibited SARS-CoV-2 replication, including bruceine A and gamabufotalin, which also demonstrated efficacy against the Delta and Omicron variants in mouse models [79]. Drug repurposing efforts also benefit from application of HTFI to project pipelines, with researchers leveraging high content imaging of infected cells against the ReFRAME drug library to identify direct-acting antivirals like nelfinavir and MK-4482, both of which demonstrated efficacy in reducing SARS-CoV-2 replication in primary human cell models and animal studies [80].

Beyond viral entry, fluorescence imaging plays a crucial role in tracking intracellular viral replication and host responses. High-throughput imaging assays have been optimized for convalescent sera screening to identify suitable donors for serum therapy, utilizing immunofluorescence and nucleocapsid-specific antibodies conserved across SARS-CoV-2 variants [81]. HTFI screening efforts have also identified monoclonal antibodies and kinase inhibitors, such as PI3K and mTOR inhibitors, that neutralize SARS-CoV-2 and block viral entry [82].

The versatility of fluorescence imaging extends beyond SARS-CoV-2. A multiplex antiviral assay was developed to simultaneously screen against multiple viruses, including Dengue, Japanese encephalitis, and yellow fever, by tagging each virus with distinct fluorescent proteins. This approach streamlines analysis and enhances the identification of broad-spectrum antivirals [83]. Similarly, a fluorescent reporter strain of severe fever with thrombocytopenia syndrome virus (SFTSV) was engineered to facilitate high-throughput screening of antiviral agents, identifying favipiravir and chloroquine as potential inhibitors [84].

Fluorescence-based approaches are also instrumental in studying the entire life cycle of hepatitis B virus (HBV). A comprehensive high-content imaging platform enabled monitoring of HBV infection, replication, and progeny infectivity, and identified inhibitors targeting various stages of the HBV life cycle, including pranlukast and fludarabine [85]. Similarly, a custom-engineered microchannel array compatible with high-resolution fluorescence imaging provided insights into influenza A and coronavirus infection kinetics, further demonstrating the adaptability of fluorescence imaging to diverse viral pathogens [86].

6.2. Screening for novel anti-pathogen drugs

Recent advancements in fluorescence imaging also highlight its utility in discovery efforts aimed at infectious diseases and antimicrobial resistance. For drug-resistant Trichomonas vaginalis, a SYBR Green-based viability assay developed by Qianqian et al. enables sensitive, high-throughput screening of potential therapeutics [87]. Similarly, Matthew V.X. et al. (2021) introduced a high-content imaging approach for Campylobacter jejuni biofilms, using TAMRA and Sytox fluorescent markers to assess biofilm integrity and extracellular polymeric substance composition [88]. Relatedly, Fabiola et al. (2024) developed a confocal microscopy-based phagocytosis assay to evaluate monoclonal antibody (mAb) activity against Neisseria gonorrhoeae [89]. This high-throughput method leverages deep learning for image analysis and supports applications across bacterial species.

In parasitic infections, Migla et al. (2022) developed reportedly the first high-content assay targeting kinetoplast DNA (kDNA) maintenance in kinetoplastid parasites, combining parasite-specific staining and automated image analysis [90]. James L. et al. (2024) advanced malaria research with a miniaturized imaging assay quantifying protein synthesis in Plasmodium berghei liver stages [91]. This assay identified novel compounds with multistage antiplasmodial activity, emphasizing the utility of fluorescence imaging in drug-resistant malaria studies.

6.3. Cancer therapeutics and screening platforms

High-content fluorescence imaging is revolutionizing cancer research by enabling detailed analysis of tumor biology, therapeutic targets, and drug safety. Ali et al. (2022) introduced MACSima Imaging Cyclic Staining (MICS), a microscopy platform capable of multiplexing over 300 protein targets in a single specimen [92]. By identifying the EpCAM/Thy1 protein pair as a target for chimeric antigen receptor (CAR) T cell therapy in ovarian cancer, MICS demonstrated its potential to guide immunotherapy while minimizing off-target toxicity. Similarly, high-content imaging has advanced drug safety profiling, with Jin et al. (2022) developing a platform to study cardiotoxicity in cancer therapies, and Yang-Chen et al. (2023) establishing a model to screen neuroprotective agents for chemotherapy-induced peripheral neuropathy [93, 94]. In melanoma, Guo et al. (2023) used imaging to identify ATM kinase as a modulator of tumor invasion, while Ningning et al. (2021) developed a fluorescent reporter to monitor O-GlcNAcylation, identifying sesamin as an inhibitor of cancer cell migration [95, 96]. Finally, Madeleine et al. (2022) presented a high-throughput imaging platform for screening radiosensitizers, identifying proflavine hemisulfate as a radiation enhancer in breast cancer models [97]. These advancements underscore the versatility of fluorescence imaging in uncovering therapeutic targets, evaluating treatment safety, and guiding personalized cancer therapies.

7. Expert opinion

High-throughput fluorescence imaging has become an indispensable tool in modern drug discovery, enabling efficient and precise interrogation of complex biological processes. By facilitating the visualization of molecular and cellular responses at unprecedented scales and speeds, HTFI supports critical applications ranging from phenotypic screening to 3D tissue modeling across drug discovery pipelines. The rise of artificial intelligence and machine learning has further enhanced the utility of HTFI, streamlining workflows by automating image analysis, improving feature extraction, and enabling predictive modeling of drug efficacy and toxicity. Looking forward, the development of interoperable AI frameworks capable of adapting to diverse imaging platforms, mitigating batch effects, and generating synthetic data to augment training datasets could further standardize and optimize HTFI pipelines. This evolution is essential for ensuring the reproducibility of results and improving the robustness of screening campaigns, ensuring that quality hits are identified early.

Standardization and validation remain critical challenges in HTFI campaigns, particularly given the variability across imaging platforms, assay designs, and data analysis protocols. Establishing universal guidelines and protocols for HTFI assays, along with public repositories for annotated datasets, will enhance reproducibility and comparability across studies. Such efforts could facilitate method validation and accelerate innovation by providing a foundation for collaborative advancements. Another promising frontier is the integration of HTFI data with other omics approaches, including genomics, transcriptomics, and proteomics. By contextualizing phenotypic insights within the broader molecular context, multiomic integration has the potential to deepen our understanding of drug mechanisms of action and disease biology. Advances in computational tools for multi-modal data integration and visualization will be pivotal for achieving these goals, enabling more comprehensive and holistic analyses.

The expansion of 3D models, including organoids, spheroids, and microfluidic organ-on-a-chip systems, represents a significant leap in improving the physiological relevance of HTFI assays. These models bridge the gap between in vitro experiments and in vivo outcomes by better recapitulating the architecture and microenvironment of native tissues. However, scaling these systems for high-throughput applications requires continued innovations in automated imaging pipelines and advanced analytical tools. Emerging technologies such as light-sheet microscopy, stroboscopic imaging flow cytometry, and real-time 4D imaging are poised to redefine the capabilities of HTFI by enabling more detailed and dynamic analyses of biological systems. Moreover, the development of novel molecular probes and biosensors tailored to specific biological targets will enhance the sensitivity and specificity of fluorescence-based assays, further expanding the applications of HTFI.

HTFI has already revolutionized drug discovery by enabling high-throughput, high-content analyses across diverse disease models and therapeutic areas. However, its continued advancement depends on addressing critical challenges in data standardization, scalability, and integration with other omics technologies. One persistent technical challenge in HTFI assays is cellular or reagent autofluorescence interference, which can significantly reduce assay sensitivity and specificity [98]. Addressing this challenge through improved probe design, spectral unmixing, and computational correction methods remains an essential area for future development. Collaborative efforts among academic, industrial, and regulatory stakeholders will be crucial for realizing the full potential of HTFI in drug discovery and development. By embracing emerging technologies and fostering innovation in assay design, data analysis, and imaging platforms, HTFI is well-positioned to drive the next generation of therapeutic development, offering new insights into complex disease mechanisms and accelerating the path toward effective treatments.

Table 1:

A summary table clearly listing specialized imaging acquisition systems and computational tools required for HTFI

Platform/Method	Instrument Required / Technology	Application Area	Reference
Mantis (Live-cell 4D imaging)	Oblique Light-sheet Fluorescence Microscope, “Shrimpy” software (open-source)	Dynamic cellular processes, viral infection tracking	Ivanov et al., 2024 [42]
FACT (Fast and Accurate Real-Time Cell Tracking)	GPU-based Segmentation & Cell Tracking Software	Real-time large-scale image segmentation	Chou et al., 2023 [46]
Fisheye Transformation Method	Deep-learning image analysis software	Single-cell phenotyping, spatial context	Toth et al., 2022 [43]
MACSima Imaging (MICS)	MACSima Cyclic Immunofluorescence Platform (Miltenyi Biotec)	Multiplexed immunophenotyping, oncology	Kinkhabwala et al., 2022 [82]
MAC-SEQ	RNA sequencing with high-content imaging	Transcriptomic & phenotypic screening	Li et al., 2023 [40]
SMAP/ProExM	Superhydrophobic Microwell Arrays & Expansion Microscopy	Super-resolution structural cellular imaging	Xie et al., 2023 [7]
Multiparametric Imaging Flow Cytometry	Sheathless microfluidic imaging flow cytometer	Sub-cellular detection, high-throughput assays	Holzner et al., 2021 [44]
Acu2Net	Deep learning segmentation algorithm	Drug screening in 3D organoids	Zhang et al., 2023 [23]

Table 2:

Overview of fluorophore probes discussed in this review showcasing their structure and target application.

Reporter	Structure	Purpose	Reference
LXMB		Near-infrared probe for β-lactamase activity	[55]
UNAG-2		Imaging of bacterial systems in anaerobic environments	[56]
SC4A-LCG		Histone deacetylase 1 activity probe	[57]
F8		Reporter substrate for cytochrome P450 3A4	[58]
BDPn2-CES		Imaging of carboxylesterase activity	[59]
Palmitoyl-CoA		Visualization of DHHC palmitoyl transferase activity	[60]
HCFP		Tracking autophagic flux	[61]

Articles Highlights Box.

• High-throughput fluorescence imaging significantly enhances drug discovery programs by enabling rapid high-content visualization of cellular responses, improving the efficiency of screening campaigns and reducing drug development timelines.

• Phenotypic screening with HTFI, with techniques like cell painting, provides unbiased identification of pathway modulation which is particularly useful for complex diseases such as cancer, neurodegeneration, and rare genetic disorders.

• Advanced 3D models and organoids, integrated with HTFI, deliver more physiologically relevant insights into drug action, overcoming limitations of traditional 2D cultures and improving predictive accuracy in preclinical testing.

• Integration of artificial intelligence and machine learning boosts HTFI’s analytical power, enabling robust, rapid, and unbiased analysis of large and complex imaging datasets, thereby accelerating drug efficacy and safety assessments.

• Newly developed fluorescence probes and biosensors allow for more sensitive, real-time monitoring of cellular and molecular dynamics, expanding capabilities for visualizing drug-target interactions and disease-related cellular events.

• Emerging automation and computational tools, such as deep-learning segmentation frameworks, address challenges related to data management, interpretation, and standardization, significantly enhancing reproducibility and throughput of HTFI assays.